Validity of self-reported assessment of Severity of Dependence Scale in Medication-Overuse Headache

2.1 Design, participants and data collection

The study was undertaken in South-eastern Norway in 2011–2013 and was a sub-study of the randomised controlled trial of brief intervention for MOH (the “BIMOH”-study) (NCT01314768) [6]. A short screening questionnaire for headache and medication use was posted to all 18–50 year old patients (10,579 out of 25,486 responded) on the patient lists of 50 participating general practitioners (GPs). As every person in Norway is listed with a personal GP, the sample is suggested to be representative of the general population. In the present study, we analysed data from participants who initially had

1. MOH (n=53)

2. headache without medication overuse (n=19)

3. chronic headache (n=25)

At the long-term follow-up, a headache expert interviewed all patients by telephone. This interview also included the five questions of the SDS. Two to three weeks after the telephone interview, the participants filled in the SDS as a part of an extensive self-report questionnaire.

The five questions of the SDS are [4];

1. Do you think your use of [headache medication] is out of control?

2. Does the prospect of missing a dose make you anxious or worried?

3. Do you worry about your use of your [headache medication]?

4. Do you wish you could stop?

5. How difficult do you find it to stop or go without your [headache medication]?

Question 1–4 are scored on a 0–3 scale, i.e. never/almost never=0, sometimes=1, often=2, always/nearly always=3. Question 5 is scored on a 0–3 scale, i.e. not difficult=0, quite difficult=1, very difficult=2, impossible=3.

2.2 Statistics

Data were described by frequencies and percentages or means and standard deviations (SD), as appropriate. Cronbach’s alpha was calculated to assess internal consistency in interview-based and self-reported SDS scales. The differences between the interview-based and the self-reported SDS sum scores were assessed by the paired samples t-test. As single SDS items are measured on ordinal scale and the differences between the interview-based and the self-reported items were reasonably symmetrically distributed, Wilcoxon matched-pair signed-rank test was applied for comparing medians. Related samples McNemar test was applied for comparison of dichotomised SDS scores. Percentage agreement between corresponding items in interview-based and self-reported SDS scale was assessed.

Pearson correlation coefficient between two methods of measurement of SDS sum scores was calculated. Agreement between the two methods was assessed by Bland-Altman analysis [12]. According to this approach, the differences between observations in the two methods on the same subjects are plotted against the mean to assess potential bias between the measurement methods. The 95% limits of agreement, defined as mean difference ±1.96 SD of the differences, are calculated and provide an interval within which 95% of differences between measurements by the two methods are expected to fall. Even though no acceptable limits could be defined a priori, the 95% limits of agreement can still be used as indicator for agreement between the methods.

4.1 Methodological discussion

The recruitment base for the study was from patient lists of representative GPs. We can never be fully sure that such a small sample is completely representative of the general population of Norway. However, for the purpose of comparing SDS in interviews with self-reports in a primary care context, we suggest the sample to be adequate. In addition, our sample is similar to samples recruited from the general population in recent studies [4], [13]. Clinical data and SDS scores were collected through interview by trained interviewers. Due to time difference between the telephone interviews and the completion of the self-reported questionnaires, and the respondents being unaware of the interview SDS scores and the inclusion of the SDS questions in a more extensive questionnaire, recall bias is likely to be minimal.

4.2 Result discussion

There was a high correlation of 0.78 between the two different methods of measurement. Despite the high correlation and the non-significant bias, the Bland-Altman analysis shows a trend that the differences between the two versions become somewhat smaller with higher SDS sum scores. Even though the differences between the methods were smallest for the part of the SDS scale most relevant for assessing MOH patients, it should be underlined that a certain underestimation may appear in the self-report form. Interestingly, question 4, “Do you wish you could stop?” was significantly differently scored and had a low percentage agreement between the two versions. The interview is by definition the gold standard, however, one may speculate whether or not this reflects an underestimation of the wish to stop in the interview or an overestimation in the self-reported version. In addition, the present results supported our previous study in suggesting that question 1 contains the least information as almost all patients (85–90%) report that they feel they have control over their medication. As previously suggested, this may suggest a lack of self-awareness of loss of control [14].

We have previously reported that SDS scores predict prognosis related to MOH and that they correlate with the type of medication overused [5], [8]. A patient with a high SDS, especially if overusing centrally acting medication, may have less chance of successful out-patient withdrawal and may potentially benefit for earlier referral for in-patient treatment or early prophylactic treatment [9]. Patients with lower scores could initially be offered medication withdrawal as out-patients with referral still an option if this fails [6], [15]. Thus, the SDS may add useful information in a clinical setting, not least in primary care. Furthermore, the self-reported SDS may be used in questionnaire-based studies of patients with MOH. Therefore, we suggest that a self-report version may be useful and yield valuable information, especially as a screening tool prior to headache consultations. The possibilities of designing web-based SDS self-assessment linked with information of MOH may be an additional advantage.